We are pleased to announce the release of v1.35.0, which includes several new features, improvements, and bug fixes. Please note that this release can contain breaking changes.
⚠️ Breaking changes
- Upgrading from versions < v1.30.0 to v1.35 is not supported directly. This version drops support for tokens generated by Deeploy < v1.30.0 in favor of a more future proof token format introduced in v1.30. When upgrading from < v1.30.0, please first upgrade to v1.34.1 and regenerate your tokens before upgrading to 1.35.0.
✨ New features
Configurable compliance insights
It is now possible to use custom compliance documentation templates, enabling better control over compliance requirements.
It is now possible to directly view container logs in the UI. This provides a more convenient and user-friendly way to monitor and troubleshoot your deployments by observing logs in real-time. Available under the prediction tab in a deployment.
HuggingFace repositories are now supported
Integrated explainers: it is now possible to deploy models with integrated explainers (e.g. Captum for Pytorch).
Visualisation of the output of integrated Captum explainers (text & image) is now possible in the UI.
Improved date selection component usability
Some default explainer images have been split into more modular packages, improving deployment time and memory usage
Improved deployment logs for restoring a deployment
Improved caching of frontend resources
Added model card to responsible Deployment checks
🐛 Bug fixes
- Resetting custom model resources could end up in an error state
- Fixed a bug with label classes in evaluation data